New firmware update given by EMILIE GILLET on Mutable Instruments forum

Official notes:
Fixed a spurious retriggering of the internal envelope when entering/leaving the speech synthesis model.
Reduced the amount of software filtering on the TIMBRE and MORPH knobs, to increase their responsiveness.
Reduced the amount of hysteresis in the quantizer responsible for converting a voltage into discrete settings (for example, for model selection with the MODEL CV input, or chord selection with the HARMONICS knob and CV input). This makes the relationship between a voltage and a specific setting more consistent when controlled by a sequencer.
The filters used for bandlimited wavetable synthesis have been improved, eliminating the amplitude drop observed for the highest notes.
The frequency range selector has a new mode: octaves. In this mode, the right button and FREQUENCY knob combo adjusts the root note, then the FREQUENCY knob transposes this root note up or down.
Pressing both buttons simultaneously for a short time enables or disables an alternative way of selecting a synthesis model: the first button selects the previous model, the second button the next one.
This new /+ control scheme gives access to an orange bank of synthesis models for pitched sounds.
1 Classic waveshapes with filter (HARMO: resonance and filter character gentle 24dB/octave CCW, harsh 12dB/octave CW, TIMBRE: filter cutoff, MORPH: waveform and sub level, OUT: LP output, AUX: 12dB/octave HP output)
2 Phase distortion and modulation (HARMO: distortion frequency, TIMBRE: distortion amount, MORPH: distortion asymmetry, OUT: carrier is sync’ed (phase distortion), AUX: carrier is freerunning (phase modulation)).
3, 4, 5 2voice, 6operator FM synth with 32 presets (HARMO: preset selection, TIMBRE: modulator(s) level, MORPH: envelope and modulation stretching/timetravel). The two voices are alternatively triggered whenever a trigger is received on the TRIG input. When the TRIG input is not patched, a single voice plays as a drone, and the MORPH knob allows timetravel along the envelopes and modulations. The LEVEL input is repurposed as a velocity control controlling loudness or timbre depending on how each preset is programmed. Each model corresponds to a different bank of 32 presets among basses/synths, keyboards/plucked strings/percussions, organs/pads/strings/brass.
6 Wave terrain synthesis with continuous interpolation between eight 2D terrains (HARMO: terrain, TIMBRE: path radius, MORPH: path offset, OUT: Direct terrain height (z), AUX: Terrain height interpreted as phase distortion (sin(y+z)).
7 String machine emulation with stereo filter and chorus (HARMO: chord, TIMBRE: chorus/filter amount, MORPH: waveform, OUT: voices 1&3 predominantly, AUX: voices 2&4 predominantly).
8 Four variable square voices for chords or arpeggios (HARMO: chord, TIMBRE: arpeggio type or chord inversion, MORPH: PW/Sync, OUT: square wave voices, AUX: NES triangle voice, TIMBRE attenuverter: envelope shape). Plug a trigger input to clock the arpeggiator.
The module can receive audio data on its TIMBRE input to load custom data into a synthesis model (only one synthesis model can be altered at a time). This new feature is available for the following models:
6operator FM: transfer of 32 custom patches in DX7 SysEx format. They will replace one of the builtin banks.
Wave terrain synthesis: transfer of a custom wave terrain.
Wavetable synthesis: transfer of 15 custom waveforms and customization of bank 4’s wave map.